smart grid testbed for demand focused energy management in … · smart grid testbed for demand...

10
arXiv:1603.06756v1 [cs.SY] 22 Mar 2016 1 Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai, Shisheng Huang, Kristin L. Wood, See Gim Kerk and Zaiyue Yang Abstract—Successful deployment of smart grids necessitates experimental validities of their state-of-the-art designs in two- way communications, real-time demand response and monitoring of consumers’ energy usage behavior. The objective is to observe consumers’ energy usage pattern and exploit this information to assist the grid in designing incentives, energy management mechanisms, and real-time demand response protocols; so as help the grid achieving lower costs and improve energy supply stability. Further, by feeding the observed information back to the consumers instantaneously, it is also possible to promote energy efficient behavior among the users. To this end, this paper performs a literature survey on smart grid testbeds around the world, and presents the main accomplishments towards realizing a smart grid testbed at the Singapore University of Technology and Design (SUTD). The testbed is able to monitor, analyze and evaluate smart grid communication network design and control mechanisms, and test the suitability of various communications networks for both residential and commercial buildings. The testbeds are deployed within the SUTD student dormitories and the main university campus to monitor and record end- user energy consumption in real-time, which will enable us to design incentives, control algorithms and real-time demand response schemes. The testbed also provides an effective channel to evaluate the needs on communication networks to support various smart grid applications. In addition, our initial results demonstrate that our testbed can provide an effective platform to identify energy wastage, and prompt the needs of a secure communications channel as the energy usage pattern can provide privacy related information on individual user. I. I NTRODUCTION The smart grid is a power network composed of intelli- gent nodes that can operate, communicate, and interact au- tonomously to efficiently deliver electricity to their consumers. It features ubiquitous interconnections of power equipments to enable two-way flow of information and electricity so as to shape the demand in order to balance the supply and demand in real-time. Such pervasive equipment interconnections neces- sitate a full-fledge communication infrastructure to leverage a fast, accurate, and reliable information flow in the smart grid. In this context, the research on different aspects of smart grid has gained significant attention in the past few years, e.g., the literatures surveyed in [1]. Although a lot has been done from theoretical perspectives, it is until recently when the actual implementation of prototypes has been given deliberate consideration. Currently, a considerable number of research W. Tushar, C. Yuen and K. L. Wood are with the Singapore University of Technology and Design (SUTD), Singapore. B. Chai is with State Grid Global Energy Interconnection Research Institute, Beijing, 102211, China. S. Huang is with the Ministry of Home Affairs, Singapore. S. G. Kerk is with the Power Automation, Singapore. Z. Yang is with Zhejiang University, China. groups are working towards establishing testbeds to validate designs and implemented protocols related to smart grid. These testbeds have various aims, scale, limitations and features; these have been summarized and presented in Table I. Most of these testbeds conduct experiments either in a lab environment (e.g., SmartGridLab, VAST, Micro Grid Lab, cyber-physical), or in isolation, in a residential (e.g.,PowerMatching City) or a commercial space (e.g., Smart Microgrid). In the testbeds surveyed, only the JEJU testbed is comprehensive enough to consider both the residential and commercial paradigm with user and grid interactions. Moreover, in spite of considerable on-going studies towards smart grid prototypes, and the massive efforts by utilities and local authorities, deployment of smart grids has met with near customer rejection 1 . Therefore, there is much impetus on designing and implementing well accepted solutions to bolster consumer-grid interactions. In this respect, this paper presents the main accomplish- ments towards a user-centric smart grid testbed design at the Singapore University of Technology and Design (SUTD). The highlight of the paper is mainly on the communication infrastructure that has been implemented in the testbed in order to provide ICT services to support a greener smart grid. The testbed is simulated in an approximate real world scenario, where the student dormitory (a 3 bedroom unit with 6 to 9 resident students) is the approximated residential space, and the faculty offices and shared meeting rooms are proxy commercial office spaces. The study focuses on both residen- tial and commercial consumers and their interaction with a central administrative body, e.g., the grid or the intelligent energy system (IES), where the IES is an entity that provides alternative energy management services to the consumers. The testbed at SUTD consists of two networks: a home area network (HAN) and a neighborhood area network (NAN). A HAN network is implemented within each residential unit or at SUTD office in order to collect different energy related data via a unified home gateway (UHG). A NAN, on the contrary, is developed to connect each HAN to the data concentrator through various communication protocols. We provide further detail discussion on both HAN and NAN networks that are implemented at SUTD in Section III. Due to the interactive nature of the system, an efficient and reliable communication infrastructure of low delay is very important. Therefore, this paper mainly discusses the various communication aspects such as broadband power line com- 1 http://www.greentechmedia.com/articles/read/illinois-rejects-amerens- smart-grid-plans1.

Upload: phungthu

Post on 29-Jul-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

arX

iv:1

603.

0675

6v1

[cs.

SY

] 22

Mar

201

61

Smart Grid Testbed for Demand Focused EnergyManagement in End User Environments

Wayes Tushar, Chau Yuen, Bo Chai, Shisheng Huang, Kristin L.Wood, See Gim Kerk and Zaiyue Yang

Abstract—Successful deployment of smart grids necessitatesexperimental validities of their state-of-the-art designs in two-way communications, real-time demand response and monitoringof consumers’ energy usage behavior. The objective is to observeconsumers’ energy usage pattern and exploit this informationto assist the grid in designing incentives, energy managementmechanisms, and real-time demand response protocols; so ashelp the grid achieving lower costs and improve energy supplystability. Further, by feeding the observed information back tothe consumers instantaneously, it is also possible to promoteenergy efficient behavior among the users. To this end, this paperperforms a literature survey on smart grid testbeds around theworld, and presents the main accomplishments towards realizinga smart grid testbed at the Singapore University of Technologyand Design (SUTD). The testbed is able to monitor, analyze andevaluate smart grid communication network design and controlmechanisms, and test the suitability of various communicationsnetworks for both residential and commercial buildings. Thetestbeds are deployed within the SUTD student dormitoriesand the main university campus to monitor and record end-user energy consumption in real-time, which will enable usto design incentives, control algorithms and real-time demandresponse schemes. The testbed also provides an effective channelto evaluate the needs on communication networks to supportvarious smart grid applications. In addition, our initial r esultsdemonstrate that our testbed can provide an effective platformto identify energy wastage, and prompt the needs of a securecommunications channel as the energy usage pattern can provideprivacy related information on individual user.

I. I NTRODUCTION

The smart grid is a power network composed of intelli-gent nodes that can operate, communicate, and interact au-tonomously to efficiently deliver electricity to their consumers.It features ubiquitous interconnections of power equipmentsto enable two-way flow of information and electricity so as toshape the demand in order to balance the supply and demandin real-time. Such pervasive equipment interconnections neces-sitate a full-fledge communication infrastructure to leveragea fast, accurate, and reliable information flow in the smartgrid. In this context, the research on different aspects of smartgrid has gained significant attention in the past few years,e.g., the literatures surveyed in [1]. Although a lot has beendone from theoretical perspectives, it is until recently when theactual implementation of prototypes has been given deliberateconsideration. Currently, a considerable number of research

W. Tushar, C. Yuen and K. L. Wood are with the Singapore Universityof Technology and Design (SUTD), Singapore.

B. Chai is with State Grid Global Energy Interconnection ResearchInstitute, Beijing, 102211, China.

S. Huang is with the Ministry of Home Affairs, Singapore.S. G. Kerk is with the Power Automation, Singapore.Z. Yang is with Zhejiang University, China.

groups are working towards establishing testbeds to validatedesigns and implemented protocols related to smart grid. Thesetestbeds have various aims, scale, limitations and features;these have been summarized and presented in Table I.

Most of these testbeds conduct experiments either in alab environment (e.g., SmartGridLab, VAST, Micro GridLab, cyber-physical), or in isolation, in a residential(e.g.,PowerMatching City) or a commercial space (e.g., SmartMicrogrid). In the testbeds surveyed, only the JEJU testbedis comprehensive enough to consider both the residentialand commercial paradigm with user and grid interactions.Moreover, in spite of considerable on-going studies towardssmart grid prototypes, and the massive efforts by utilitiesandlocal authorities, deployment of smart grids has met withnear customer rejection1. Therefore, there is much impetus ondesigning and implementing well accepted solutions to bolsterconsumer-grid interactions.

In this respect, this paper presents the main accomplish-ments towards a user-centric smart grid testbed design atthe Singapore University of Technology and Design (SUTD).The highlight of the paper is mainly on the communicationinfrastructure that has been implemented in the testbed inorder to provide ICT services to support a greener smartgrid. The testbed is simulated in an approximate real worldscenario, where the student dormitory (a 3 bedroom unit with6 to 9 resident students) is the approximated residential space,and the faculty offices and shared meeting rooms are proxycommercial office spaces. The study focuses on both residen-tial and commercial consumers and their interaction with acentral administrative body, e.g., the grid or the intelligentenergy system (IES), where the IES is an entity that providesalternative energy management services to the consumers.

The testbed at SUTD consists of two networks: a home areanetwork (HAN) and a neighborhood area network (NAN). AHAN network is implemented within each residential unit orat SUTD office in order to collect different energy related datavia a unified home gateway (UHG). A NAN, on the contrary,is developed to connect each HAN to the data concentratorthrough various communication protocols. We provide furtherdetail discussion on both HAN and NAN networks that areimplemented at SUTD in Section III.

Due to the interactive nature of the system, an efficientand reliable communication infrastructure of low delay is veryimportant. Therefore, this paper mainly discusses the variouscommunication aspects such as broadband power line com-

1http://www.greentechmedia.com/articles/read/illinois-rejects-amerens-smart-grid-plans1.

Page 2: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

2

munication cables (BPL), TV white space (TVWS), HAN andNAN of the testbed at SUTD. Benefits that could be attributedand examined in this system include lowering of operationalcosts, increasing ancillary electricity options and better peakdemand management. We would also design and examinesuitable incentives, which would encourage consumers toadopt these demand response mechanisms for participation inthe market by exploiting the use of the testbed.

II. T ESTBEDCOMPONENTS, FEATURES AND CHALLENGES

The smart grid is envisioned as a future power networkwith hundreds of millions of endpoints, where each endpointmay generate, sense, compute, communicate and actuate. De-ployment of such complex systems necessitates sophisticatedvalidation prior installation, e.g., by establishing smart gridtestbeds. However, the rapid fluctuation in power supplyand demand, voltage and frequency, and active consumers’interaction with the grid make the practical realization oftestbeds extremely challenging. To this end, we classify thesuitable technologies and features of such testbeds into threecategories: 1) Hardware based components, 2) Software basedelements, and 3) Other features. We give a brief descriptionofdifferent technologies and elements within each of the definedcategories and their impact on the design of a communicationnetwork as follows.

A. Hardware based components

1) Two-way communication network:In a smart grid, alarge number of sensors, actuators, and communication devicesare deployed at generators, transformers, storage devices,electric vehicles, smart appliances, along power lines andin the distributed energy resources. The optimal control andmanagement of these nodes in real-time is essential for thesuccessful deployment of smart grids [2], which gives rise tothe need for fast and reliable two-way communication infras-tructures. For example, depending on the online requirementsto response, reserve power is required in the smart grid in caseof an unexpected outage of a scheduled energy resource, andrequired response has to be deployed within30 seconds2 .

2) Advanced metering infrastructure (AMI):The AMI mea-sures, collects and analyzes the energy usage, and communi-cates with metering devices either on requests or on a schedule.There is an increasing concern regarding their costs, securityand privacy effects, and the remote controllable “kill switch”included in them. A scalable secured MAC protocol becomescritical for the AMI system.

3) HAN and NAN: In accordance with [3], we organizethe smart grid network into a HAN and NAN. A HANis the core component of establishing a residential energymanagement system as part of a demand response management(DRM) scheme in a smart grid. Establishing the HAN isrequired to maintain an internet-of-things protocol for differentsensors/actuators that run on different physical layer commu-nication protocols within a home. This consequently imposes

2Reserve power can be split into primary (less than30 seconds),secondary (less than15 minutes) and tertiary (15 minutes), depending onrequired response times to monitor and control.

the necessity of devising gateways/controllers/central commu-nicators, which are capable of handling these miscellaneouscommunication techniques. Depending on the business model,there could be multiple NANs. For example, AMI is setup andmaintained by the grid, and is connected to one NAN; whilethird party IES (e.g. Google NEST) could be on another NANthat provides additional services such as home automation,energy management, or security services.

4) Renewables supply:The generation from renewablesources such as wind and solar is intermittent and unpre-dictable [4], [5]. Hence, how to design the control architec-tures, algorithms, and market mechanisms to balance supplyand demand, support voltage, and regulate frequency in real-time are the key challenges in the presence of these volatilegreen energy supplies.

5) Smart power electronics:Smart components such asDC/AC inverters, flexible AC transmission systems and smartswitches will be increasingly deployed on various componentssuch as the transmission systems, transformers and powergenerators in the smart grid. Hence, there are needs of so-lutions that will enable the design of power electronics andcontrol appliances to maximize reliability, efficiency andcostreduction for the power grid.

6) Energy storage system:An energy storage system storeselectricity when the demand is low and provides the grid/IESwith electricity when the demand is high. Energy storagesystems are becoming an essential part of smart grids due tothe increasing integration of intermittent renewables andthedevelopment of micro grids [6]. Innovative and user-centricsolutions are required to determine the type and size of energystorage according to the application, to determine the optimalstorage location and reduce the cost of storage devices.

7) Electric vehicles (EVs):EVs can act as mobile storagedevices and assist the grid to fulfil various energy managementobjectives through grid-to-vehicle (G2V) and vehicle-to-grid(V2G) settings. Key challenges are: addressing issues likequick response time, power sharing, EV charging protocoldesign and exploiting the use of EVs storage devices forpower compensation [4], and voltage and frequency regulationservices [7].

B. Software based elements

1) DRM: The two-way flow of information and electricityin smart grids establishes the foundation of DRM; which is infact the change of electricity usage patterns by the end-usersin response to the incentives or changes in price of electricityover time. DRM can be accomplished in either a centralized ora distributed fashion [8]. A distributed DRM algorithm greatlyreduces the amount of information that has to be transmittedwhen compared to a centralized DRM algorithm. Therefore, asmart grid testbed needs to have the capability to investigateboth DRM schemes, if possible, in order to examine theireffectiveness on the target load management.

2) Dynamic pricing: Dynamic pricing is critical for effec-tive DRM in smart grids as a driver for behavior change [9].The pricing scheme needs to be beneficial to the grid interms of reduction in operational cost, energy peak shaving

Page 3: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

3

TABLE I: Attributes of Different Smart Grid Testbeds.

Testbeds Set-up Adopted Communication Technology Main Innovations

Components and

Features

(From Section II)

SmartGridLab

(Washington State University) In a Lab environment.

IEEE 802.15.4 wireless technology is being used, which

is the basis for ZigBee, ISA100.11a, WirelessHART and

MiWi.

Intelligent power switch control and real-time demand response and

self-healing, energy storage and flow balance algorithms.

A.1, A.2, A.6, B.1,

B.4

Cyber-physical

(University of Illinois at Urbana-

Champaign)

In a Lab environment. Have not used any particular communication technology. A multi-platform protocol dissector library, an AMI visualization

framework and Advanced Metering Infrastructure (AMI) security. A.1, A.2, B.4

Smart Microgrid

(British Columbia Institute of

Technology)

Enable optimal operation of a

university campus power

system.

IEC-61850 based bi-directional communication link is

used at the BCIT Intelligent micro-grid.

Control strategies for efficient integration and operation of on-site

DER units, energy storage and implementation of fault detection,

isolation and restoration protocols.

A.1, A.4, A.5, A.6

VAST

(Georgia Institute of Technology)

In a Lab environment.

A virtual communication link is used that allows the

testbed to implement any type of communication

protocol required. Reprogramming the system using

LABVIEW does the implementation of different

protocols.

Virtual communication link and Programmable inverter control,

identification of false or incomplete data, and high latency

coordination of response to events such as a fault or

coupling/decoupling to an external grid.

A.1, A.4, A.5, B.4

PowerMatching City

(The Netherlands)

Demonstration consists of 25

households in the city of

interest.

Data communication between EV and the grid of

PowerMatchingCity is done through VPN router over

UMTS. Communication between the households and the

central servers is provided by VPN network by a

separate ADSL connection.

A generic coordination mechanism, comfort control mechanism, and

monitoring. A.1, A.3, C.3

Smart Building Platform

(International Hellenic University

(IHU), Greece)

Testbeds are set up at IHU

campus.

Open and encrypted ZigBee, Z-wave, Wi-Fi and Radio

Frequency (RF)

Energy monitoring and management in a smart building,

environmental monitoring, demand response, web development and

smartphone application for control energy usage.

A.1, A.2, A.3, B.1,

C.3

JEJU testbed

(Korea)

Testbeds are set up for both

residential and commercial

spaces at JEJU Island in Korea.

Customer domain- SEP, ZigBee, WiFi, PLC;

Transmission and Distribution: IEC 61850, DNP 3.0;

Operations Domain IEC 61968/61970; AMR; Public and

Private Internet; and Cellular based (CDMA, WCDMA),

WiMAX (IEEE 802.16)

Power transmission using IT with solar and wind power, energy

storage, optimizing energy usage by utilizing new and renewable

energy sources, collecting real time data on energy usage and demand

to limit the unnecessary use of electricity.

A.1, A.2, A.3, B.1,

A.4, A.5, A.6, A.7,

B.2, C.3

National Smart Grid Lab

(Norway)

Testbeds are set up to offer

realistic physical model grids

and interface with generation

and loads.

Fiber-to-home communication system.

Identification of information security threats, energy control in smart

house, use of wind, solar and bio-fuel as energy sources, monitor load

profiles and electricity production in zero energy house and

determining user flexibility.

A.3, A.4, B.4, C.3

Newcastle University testbed

(UK)

Trial of new storage

technologies to efficiently and

sustainably deliver energy

across the UK electricity grid.

Have not used any specific communication technology.

Efficient use of energy storage, integration of distributed generation

and customer engagement, charging of electric vehicle and demand

response management.

A.4, A.6, A.7, B.1,

C.3

Smart grid testbed

(SUTD, Singapore)

Experimental set up includes

both student dormitories and

main university campus.

For communication within a home area network: UHG

to communicate and control different household

appliances via different communication protocols: Wi-

Fi, ZigBee, Z-wave, Bluetooth

For communication from cloud to the home or smart

meter: broadband PLC, TV White Space

Monitoring and recording real-time energy usage data to encourage

and control the energy usage of consumers, designing and evaluating

communication technology for end user environment, centralized and

distributed demand response in real-time, designing incentives for

consumers and dynamic pricing.

A.1, A.2, A.3, B.1,

B.2, B.4, C.1, C.2,

C.3

and valley filling. It also needs be economically attractivetoconsumers by reducing their electricity bills and should notcause significant inconvenience to them for changing theirenergy consumption behavior, i.e., for scheduling or throttling.Securing the pricing information is important to protect againstmalicious attacks, e.g., an outdated pricing information isinjected to destabilize the smart grid.

3) Outage management:In a time of crisis, the capacityfor fault detection and self-healing with minimal restorationtime and improved efficiency, i.e., outage management [10] isimportant. Hence, with real-time insight, the smart grid testbedshould possess automatic key processes to easily locate androute power around affected spots to reduce unnecessary truckrolls and save costs.

4) Security and privacy:Cyber security is a pivotal chal-lenge for smart grid. In fact, cyber-based threats to criticalinfrastructure are real and increasing in frequency. However,the testing of potential threats are challenging due to thegeneral lack of defined methodologies and prescribed waysto quantify security combined with the constantly evolvingthreat landscape. Moreover, as the participation of consumersin smart grid becomes more prevalent, privacy information,e.g., their home occupancy, will be more vulnerable. Hence,applied methods and rigorous privacy and security assessmentsare necessary for complex systems where heterogeneous com-ponents and their users regularly interact with each other andthe IES.

C. Other features

1) Ancillary Services:Regulation power allows for the gridto manage second-to-second fluctuations from forecasted de-mand and reserve markets enable electricity providers to haveinstant-on solution to kick in when power delivery problemsemerge. Theoretically, smart grids are able to provide suchservices through DRM [11]. Therefore, smart grid solutionsneed to enable the grid providers to combine the DRMschemes with energy storage solutions to make the grid systemmore reliable.

2) Scalability: A smart grid system may consists of mil-lions of active nodes, which need to be managed in real-time. Accordingly, optimal economic mechanisms and busi-ness models are required for engendering desired global out-comes. However, it is extremely difficult to maintain such alarge scale and complex cyber-physical system in real-time.Besides, features like automatic fault detection, self-healing,autonomous distributed demand-side management and disastermanagement make the conservation more strenuous. Hence,innovations are required to improve the scalability of suchahuge system without affecting any of its features. For instance,instead of using different gateways for communicating withdifferent equipment in a home (where each devices may runon different communication protocols), devising a single UHGcould make a HAN considerably scalable. A scalable MACprotocol would be interesting to explore in dense residentialareas, where over thousands of machine-type devices such assmart meters communicate at the same time [12].

Page 4: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

4

Fig. 1: Overview of SUTD testbed consisting of NAN and HAN.

3) User acceptability and participation:One of the keychallenges for the success of smart grids is to be able to clearlyelucidate the benefits for the consumers [13]. One solutioncould be a clear user friendly interface (e.g., smart phone apps)to demonstrate the ability to visualise electricity usage andreal-time electricity prices, so as to ease energy savings forthe consumers through smarter load shifting.

It can be noticed that advanced ICT technologies andcommunications networks play important roles in variousapplications, features, and even in the acceptance of a smartgrid. To this end, we now discuss the main features of thetestbed at SUTD in the next section.

III. SUTD TESTBEDOVERVIEW

The aims of the SUTD smart grid testbed are to developinnovative technical approaches and incentives for smart en-ergy management, through the integration of commercial spacestrategies, residential space strategies, and the design anddevelopment of supporting technologies. Hence, testing thesuitability of various communications technologies for smart

grid so as to address some of the challenges mentioned inSection II is especially critical. The interaction betweentheconsumers and the grid/IES has been prioritized in this work,which leads us to divide the whole smart grid testbed into twonetworks: 1) NAN and 2) HAN. The IES set up the HAN,which is connected to the cloud directly though broadbandinternet, while the grid will set up the NAN that connects thesmart meters. However, if the IES is provided by the grid, thenthe HAN and the UHG would certainly be connected underthe same network. To keep the structure general, we considerthem to be separated in the SUTD testbed. To this end, whatfollows are the brief descriptions of the NAN and HAN, andthe associated technologies used in the SUTD testbed.

A. NAN

The NAN of the testbed is composed of a number of HANunits from either the SUTD dormitories or the universitymain campus and the cloud. In our setup, we employ twotechnologies for NAN: one is BPL and TVWS for smartmeters setup by the grid, and another is the traditional fiber

Page 5: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

5

TABLE II: The technical specifications of TVWS network deployed at SUTDcampus.

/ cable internet, where all the UHGs are connected via Wi-Fi access points, which could be setup by a 3rd party IES.In Fig. 1, we show a schematic diagram of how the NANis set up within the SUTD testbed. Now, we provide a briefdescription of the used communications technologies, i.e., BPLand TVWS, in the following subsections.

1) BPL: In the testbed, the communication between thesmart meter outside each residential unit and the data con-centrator at the top of the apartment building is based onBPL. This choice of communication between them is simpleand obvious as no additional cabling is needed in connectingall the units within an apartment block. It provides superiorperformance across thick walls as compared to any othertypes of communications and simplified the entire networkmanagement of smart meters.

2) TVWS: At SUTD, TVWS is used for linking the dataconcentrator from every apartment block to the base-stationbefore data is uploaded to the cloud. In Table II we showthe detail specification of TVWS that we have used at SUTD.Currently, Infocomm Development Authority (IDA) is settingup trials and the standardization of TVWS in Singapore.Please note that TVWS is a freely available spectrum and thusprovides a great opportunity for many new small and virtualoperators to setup their networks for M2M applications at lowcost. In this testbed, building a NAN on TVWS spectrumalso provides more operational flexibility that can easily bescaled up by adding additional Base Stations (BSs). Therefore,for NAN wireless access provides good solution and widecoverage without much installation cost. Furthermore, TVWS(unlicensed band) can greatly reduce the cost, as compared tolicensed band and fixed access.

Our testbed based on BPL and TVWS provides a dedicatedand separated network from the Internet cabling, which isalso used for web surfing or video streaming. It provides anopportunity to test the network delay and reliability for smartgrid applications, and can be compared against the HAN thatwill be discussed next.

B. HAN

While all participating units from the dormitories and thecampus form the NAN together, each unit is equipped with its

own HAN inside the house as shown in Fig. 2. A UHG is thecentral point of attention of a HAN, through which the IEScan monitor the energy usage pattern or control each of theequipments of the unit within the network. Here, on one hand,the gateway adopts two-way communication protocols such asZigBee, Z-wave and Bluetooth to connect with equipments andsensors in the house. On the other hand, it is connected to theIES through Wi-Fi for sending the monitored information tothe IES.

1) UHG: A novel aspect of the SUTD testbed is theintroduction of a UHG in each of the HAN unit of bothresidential and commercial space areas, built on a RaspberryPi computer (Model-B Rev 1) as shown in Fig. 2. Please notethat the HAN network consists of different type of smart plugand sensors that may need to communicate with each other. Inthis context, Universal home gateway (UHG) is developed andused in the testbed to facilitate such heterogeneous communi-cation between different sensors and smart plugs. The UHGis capable of handling a number of communication protocolsincluding Z-wave, ZigBee and Bluetooth for communicatingwith different smart devices such as smart plugs and varioussensors. The UHG directly transmits the monitored data froma HAN unit to the IES through Wi-Fi. The challenge of sucha UHG is that it needs to communicate with devices (e.g.smart-plug / sensors) of various protocols. The delay betweenthe commands from the IES to the UHG and from the UHGto the end devices could also pose challenges depending onthe applications. It however provides a scalable solution thatenables thousands or even ten thousands of households in thesystem. In the SUTD testbed, we have implemented RestfulHTTP and XMPP protocol for the communication betweenthe UHG and the cloud of the IES. Hence, the UHG servesas a gateway for many non-Internet Protocol based sensors toconnect to the Internet.

2) ZigBee: ZigBee is a low-cost and low-power wirelessmesh network specification, which is built on the top of theIEEE 802.15.4 standard for low-rate wireless personal areanetworks (WPANs), and operates in the ISM frequency band2.4 GHz. ZigBee is essentially a non-IP based communicationprotocol and thus suitable for communication in testbeds setup in a non-residential setting. This is due to the fact thatin SUTD campus (which is also the same as many othercommercial/industrial network), the Wi-Fi network has verystrict security, where the password is changed on regular basis(e.g., in every three months at SUTD campus). However, it issignificantly difficult to change the password of every sensorin every three months time. Therefore, non-IP based solutionis required and we use ZigBee in the MPN to serve thatpurpose. Besides, for a multi-hop extension considering theoffice scenario, we require a protocol that can support multi-hop to extend the coverage to reduce the needs of a gateway(within a house, however, single hope with UHG is usuallygood enough). Furthermore, ZigBee provides security (e.g.the ZigBee provides basic security where only nodes of thesame network id can join in the network) and the simplicityof having two-way communication facilities.

The multi-purpose node (MPN) designed in SUTD, asshown in Fig. 2, consists of a number of low power sensors and

Page 6: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

6

Fig. 2: Home Area Network.

actuators. It is equipped with a motion detector, a temperatureand humidity sensor, a noise sensor and a Lux sensor. Itcommunicates with the UHG through its Xbee Chip, whichis a ZigBee communication module, to send the monitored in-formation. The MPN is capable of controlling devices throughits actuators like an IR Blaster (to control the air conditioningsystem (ACS)) and potentiometer (to control the LED lightpower supply). The MPN is driven by an Arduino Fio microcontroller, which is installed within the MPN. In every hostelunit in SUTD, each HAN consists of 4 MPNs, one for eachbedroom, and one for the living room. While in the office,there are 20 MPNs (one for each faculty office) in one sectionof an office block connected through multiple ZigBee relaysto a single UHG. The UHG also fulfils an important role lateron, when distributed DRM is implemented, such that usersoccupancy and electric appliance usage info can be storedlocally without sending to the IES to protect users’ privacy.

3) Z-wave: Z-wave is a wireless communication protocolaround900 MHz that uses a low power technology for homeautomation. For instance, to monitor and remotely control theenergy consumption of different equipments in the HAN, wehave used a Z-wave smart-plug for each of the devices in theunit. In the current set-up, electric appliances to be monitoredare connected to power outlets through Z-wave smart-plugs,whereby each smart-plug is connected to the UHG through aRaZberry module. The plug is able to monitor the amountof energy consumed by the connected device in real-timeand instantaneously send that information to the UHG via Z-

wave communication. Further, it also has a remote actuationcapability e.g., delaying an electric appliances due to peak inenergy demand. Note that the switching signals of smart-plugsare initiated through either the IES (for centralized control) orthe UHG (for distributed control).

4) Bluetooth: We use a Bluetooth Low Energy (BLE)sensor based on Texas Instrument’s (TI’s) CC2541 MCU tomonitor the temperature, humidity and pressure level of itssurroundings, and to communicate with the UHG via blue-tooth. The BLE sensor is also equipped with an accelerometer,a gyroscope and a magnetometer that provide the UHG withinformation like acceleration, orientation and magnetization ofan object respectively. In the course of the experiments, wewill distribute such TI sensors among the students so that theycan design new Internet-of-Things product that can connecttoour UHG and provide new green energy services in smarthomes.

5) Mobile application: We have developed a mobile ap-plication3 to install in students’ and faculties’ cellphones toengage them in the energy management experiments at SUTD.For example, they can remotely switch on and off the smartplug or the ACS, or monitor the energy usage collected by thetestbed. Dynamic pricing information can also be pushed tothem over the application.

3The figure is not given due to constraint on the number of figure.

Page 7: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

7

C. Challenges

During the experiments in the testbed, we face a number ofchallenges including communication delay, subject recruitmentfor case studies, modeling non-homogeneous scenarios andinvestigating the mismatch between the theoretical results andthe outcomes from the experiments.

First, communication delay is one of the challenges thatwe have encountered while running the experiments. Weunderstand that there is a trade-off between the sampling rateand communication delay, and the delay can be reduced byincreasing the rate of sampling. Delays can also be causedby the time required for data savings and retrieval at theserver. For example, if a server needs to retrieve data fromlarge number of sensors at the same time, it may incur somedelays [14]. However, one potential way to reduce such delayis to efficiently design the database and use more customizedscript for increasing the processing speed. Nevertheless,suchdelay is beyond the scope of this paper. Now considering theresponse times of primary, secondary and tertiary reserves,our implemented centralized monitor and control scheme cancomfortably use for secondary (less than 15 minutes) andtertiary reserve (15 minutes). However, it might not be suitablefor primary reserve requiring very fast response time, e.g., lessthan one minute. Implementing distributed control insteadofcentralized and thus enable each node under observation torespond very fast, if necessary, could be a potential way toresolve this issue.

Second challenge is to recruit the subjects for experimentswho will allow us to install sensors and communicationmodules in their rooms for monitoring and control purposes.Although the experiments are completely Institutional ReviewBoard (IRB) approved and are designed to conduct in anacademic environment, it is hard to find participants for theexperiments. This is due to the fact that the participants areconcerned about their privacy and worry about the inconve-nience that may create due to energy management. Further,students lifestyle is significantly different from the lifestyle oftypical residents. In this context, as a token of encouragementfor participation, we have provided monetary incentives toeach of the participants at the end of the study.

After the rooms were chosen to set up the testbed, thethird challenge was to interpret the readings from the sensorsinstalled within the rooms. This is due to the fact that eachuser has different preferences and hence sets the sensors atdifferent locations of the rooms. For instance, some may keepthe sensors near the window whereby some prefer to keepthem far from the windows. Similarly, some prefer to alwayskeep their window curtains open whereas some prefer them toremain close most of the time. As a consequence, the readingson temperature, light, noise, and motion from the sensors havethe possibility to be very different although they may representthe same environment. Hence, there is a need to consider suchbehavioral differences in our experiments, and we have triedto differentiate the sensors data based on the context. Forexample, we perform some sort of learning before interpretingthe readings from sensors.

Finally, we find it difficult to match the theoretical load

consumption model with the model that we derive from theexperiments. For example, let us consider the case of airconditioning systems (ACSs). Although there are many studieson the energy consumptions of ACS in the literature, it isextremely difficult to find a study that has used a setup assame as the one we use, e.g., in terms of ACS type, room sizeand weather. Therefore, algorithms that are designed basedonthe theoretical model may not behave as expected in practicalscenarios. As a result, additional steps are needed to builda model based on practical system setup, and building suchmodel could be time consuming and very customized to apractical setting.

After discussing various challenges, now we will discussthe experiments that are currently (or will be) implementedatSUTD in the next section to give a glimpse of the competenceof this extensive testbed setup.

IV. ON-GOING EXPERIMENTS

Most of the experiments using the testbed are deliberatelyrelated to electricity management either through direct controlof electricity from the IES or by designing incentives that willencourage behavioral modification towards efficient energyuse. Our experiments at the SUTD testbed are conducted1) within the student hostels, and 2) at the SUTD campus.Based on their electricity usage pattern, the SUTD campusis considered as the commercial space in our experiments,whereby student hostels are thought of as residences.

A. SUTD student hostels

The energy consumed by residences differ based on theset of home appliances, standard of living, climate, socialawareness and residence type. We use our testbed to developsolutions for residential energy management to attain differentobjectives.

1) Ancillary electricity markets:In the SUTD testbed, weare interested in determining the potential of residentialsmartgrid participation in ancillary markets. Real-time ancillarymarkets allow participants to supply both regulation andreserve power. Reserve power is generation capacity that isre-quired during an unexpected outage of a scheduled generatingplant and we are concerned with whether the residential userscan provide the grid with either primary or secondary reservepower. In most systems, discretionary loads can voluntarilyparticipate in these markets as interruptible loads, willing to“load-shed” in an exceptional event; however, current marketrules only allow participation of big loads. In our system, theIES can aggregate smaller distributed loads and participatein these markets as a unified entity. Although theoreticallypossible, the practical implementation with consideration ofpractical limitations have not been fully examined in fullscale testbeds, especially the communication constraints. Withthe SUTD testbed, we will verify its feasibility, and therequirements on the communication network in supportingsuch an application.

Page 8: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

8

En

ergy

con

sum

pti

on

En

ergy

con

sum

pti

on

En

ergy

con

sum

pti

on

Time

Time

Time

Fig. 3: Demonstration of the real-time energy consumption by different homeappliances as obtained from the deployed testbed. The horizontal axis referto different time of the day whereby the vertical axis represent the energyconsumption by different appliances.

2) Dynamic pricing: Using the implemented testbed, wealso focus on designing dynamic pricing schemes for resi-dential users. Different design objectives such as penetrationrates [8], flexibility thresholds and flexibility constraints willbe considered. As an initial step towards this development,amobile application is being developed for the residential usersat SUTD that can periodically notify them about their real-time energy consumption amount and the associated price perunit of energy. We will use such dynamic pricing informationto achieve certain DRM goals and to promote energy savingawareness. The smart meter is installed in parallel with originalmeter, such that the smart meter can be used to simulate theelectricity price based on dynamic pricing, while the originalmeter by the grid can continue provide meter readings foractual electricity billing.

3) DRM under communication impairment:Real-timeDRM can be affected significantly by the quality of thecommunication network of the smart grid. In this context,we are conducting experiments to observe the effect of thestatus of communication network on DRM. For instance, ifthe communication network is congested, it would result inpacket losses or delay. We are interested in investigating howthis congestion of the network affects the DRM in terms ofdelay and consequently the cost, reduction of peak load andenergy savings.

4) DRM design based on preferences and device character-istics: The installed smart-plugs and MPNs in each HAN unitat the SUTD student hostel enables real-time monitoring ofenergy consumption of different appliances as shown in Fig 3.This provides insights on the students personal preferencesand allows to feed the energy consumption data back to theconsumers in real-time, which along with the incentive basedDRM policies would enable the users to modify their usagepattern towards more efficient energy management.

Remark 1. It is important to note that our testbed is to

0 3 6 9 12 15 18 21 2415

20

25

30

35

40

Time (Hours)

Tot

al p

ower

(kW

)

Total load profile wihout DRMBase load profileTotal load profile with DRMPeak limit

Fig. 4: Illustration of energy management scheme in the10 hostel units inorder to reduce peak load.

monitor the energy usage. However, each appliance has aunique characteristic profile as shown in Fig. 3, which can beused to identify what appliance that is, and in return, speculatethe activity and number of users. This prompts the need forsecure communication.

Now to demonstrate how the proposed testbed can be usedto perform energy management under practical communicationimpairment, we run experiment for peak load shaving inour testbed. To do so, we have considered10 home unitswhere each unit is equipped with two flexible loads (weuse lights as flexible loads that provide on/off control). Wemodel the daily base load according to the reports of NationalElectricity Market of Singapore (NEMS) and scale accordingto the energy rating of light bulbs such that we can obtain asignificant percentage of flexible load. We assume a thresholdfor total maximum allowable load consumption by all theunits and design an algorithm to control the flexible load ofeach home units in order to keep the energy consumptionalways below the threshold. The algorithm is based on acentralized control scheme in which the energy managementservice provider conduct demand management by switchingoff several appliances of the users to maintain the totalpower demands below the given threshold. As designed, theenergy management service provider considered not only theengagement of users in the demand response but also theinconvenience of users when the demand response protocol isconducted. We find that the demand response can be affectedby communication delay. The detail of the algorithm andassumption of the experiment can be found in [15].

The results from the experiment are demonstrated in Fig. 4.In Fig. 4, the green zone and blue zone of the figure denotethe load profile with and without demand management respec-tively whereby grey zone indicate the base load and red lineshow the maximum allowable peak demand threshold (whichis 33 kW in the considered case). Now, as demonstrated inFig. 4, when the total demand is under peak limit, there is noneed to do any controlling and hence no difference is observedbetween green zone and the blue line. However, once thetotal demand exceed the peak limit, the algorithm is executedand thus controls the flexible loads in each home unit andturns some of them off to reduce demand load promptly as

Page 9: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

9

TABLE III: Demonstration of how the monitoring of users occupancy andenergy use of lights and ACS in eight office rooms at SUTD are done throughthe implemented testbed in order to reduce energy wastage. The total wastageis calculated in terms of both kWh and SGD.

indicated by the blue line. It is important to note that thealgorithm continuously run in the backend and monitors forsituations when the total demand may exceed the threshold.Once such situation arises, the algorithm switches off someof the appliances to keep the demand within the threshold.However, the algorithm, as it is designed, considers not onlythe demand control but also the associated inconveniencethat the users may experience for such control. Therefore,control of appliances is always kept at as low as possible inmaintaining the demand. Furthermore, while controlling someappliances for handling excess demand, there could be newloads switched on by the users that can also contribute to theoverall demand. As a consequence, there is no sudden decreaseis observed in the blue line as can be seen from Fig. 4. Thus,the result in Fig. 4 clearly shows that our developed testbedis effective to preform energy management applications.

B. SUTD campus

We are also interested in studying energy management incommercial buildings with a view to reduce energy costs forall stakeholders. To this end, we are currently conducting thefollowing experiments at the SUTD campus.

1) DRM in office environments:One of our ongoing experi-ments is to implement DRM schemes for offices that considersthe real-time control of ACSs. The objectives are mainlytwo-fold: 1) real-time thermostat control to manage peakdemand, and 2) optimal management of ACS demand underdynamic pricing for energy cost management. To supportthis, we are developing energy consumption models and real-time control algorithms for ACSs. The testbed allows us tomonitor and collect the data on consumed power by the ACSsunder different permutations of indoor and outdoor conditions.These are pivotal for algorithm development and ACS energyconsumption modeling.

2) Meeting scheduling:In offices, usually there exists alarge number of meeting rooms that are used only occasionally.Hence, the scheduling of their usage is a possible way toshape the energy consumption of commercial buildings. Weformulate a comprehensive study, where for a given set ofmeeting room requests over a fixed time period and a setof meeting rooms in a dynamic pricing market, the meetingscheduler finds a feasible routine that minimizes the totalenergy consumption (and/or cost). The SUTD testbed is usedfor obtaining the real-data, and to test the scheduler in order

to verify the effectiveness of the optimal meeting schedulingprotocol.

Now, in order to save energy and associated costs ofindividual office rooms, we experiment to identify the potentialwastage in an office environment. To do so, we have deployeda number of sensors, i.e., the MPN of Fig. 2 in each of theoffices. For instance, if the sensors do not detect any motionor noise in a room for a predefined period of time, we canassume that there is no one inside the room. A record of suchobservation is shown in Table III, which shows the duration ofturned-on lights and ACSs of eight different office rooms whenthere are no occupants. According to Table III, the energywastage in most of the considered rooms is significant inabsence of the occupants, which is on average6.376 kWhand232.025 kWh per room from lights and ACS respectively.Now, if we assume that the average wastage is the samefor all 200 office rooms of a typical medium size universitycampus, the total energy that wasted during the consideredtime duration of 54 days can be estimated as47, 680.2 kWh.Now considering the electricity rate in Singapore, which iscalculated via Home Electricity Audit Form available online4,this wastage can be translated into a total of around11, 100

SGD from the considered 200 rooms during a period of54

days. Therefore, if the proposed testbed can be used in officesfor the considered setting, about11, 100 SGD can be savedfrom wastage. The saving will be even more if the electricityprice is dynamic, as the office hour is typically overlap withday time peak period. One can then perform on/off control tothe ACSs (such as in Fig. 4) to achieve energy saving or peakload shaving.

V. CONCLUSION

In this paper, we have discussed some aspects on the devel-opment of the smart grid testbed at the Singapore Universityof Technology and Design (SUTD). These testbeds withinthe university campus have been setup to approximate bothresidential and commercial spaces. We have discussed thegeneral components, features and related challenges of sucha testbed implementation with emphasis on the creation ofthe communication system architecture. Ongoing experimentsat the testbed include the harnessing of detailed data streamsfrom the testbed for different customized energy managementprograms such as demand response. The flexibility and ex-tensivity of the deployed testbed allows for the research teamto explore and implement effective and practical communica-tions technologies and smart grid applications, which couldultimately increase the acceptance and adoption rate of thesesystems.

ACKNOWLEDGEMENTS

This work is supported in part by the Singapore Universityof Technology and Design through the Energy InnovationResearch Program Singapore under Grant NRF2012EWT-EIRP002-045, in part by the SUTD-MIT International Design

4The cost can be calculated by using theHome Electricity Au-dit Form available online for public use in https://services. spser-vices.sg/electaudit frameset.asp.

Page 10: Smart Grid Testbed for Demand Focused Energy Management in … · Smart Grid Testbed for Demand Focused Energy Management in End User Environments Wayes Tushar, Chau Yuen, Bo Chai,

10

Center, Singapore under Grant IDG31500106 and in part byNSFC-61550110244.

REFERENCES

[1] X. Fang, S. Misra, G. Xue, and D. Yang, “Smart grid - The newandimproved power grid: A survey,”IEEE Commun. Surveys Tuts., vol. 14,no. 4, pp. 944–980, Oct 2012.

[2] S. Maharjan, Q. Zhu, Y. Zhang, S. Gjessing, and T. Basar,“Dependabledemand response management in the smart grid: A Stackelberggameapproach,”IEEE Trans. Smart Grid, vol. 4, no. 1, pp. 120–132, Mar2013.

[3] R. Yu, Y. Zhang, S. Gjessing, C. Yuen, S. Xie, and M. Guizani,“Cognitive radio based hierarchical communications infrastructure forsmart grid,” IEEE Netw., vol. 25, no. 5, pp. 6–14, Oct 2011.

[4] W. Tushar, C. Yuen, S. Huang, D. Smith, and H. V. Poor, “Costminimization of charging stations with photovoltaic: An approach withEV classification,”IEEE Trans. Intell. Transp. Syst., vol. 17, no. 1, pp.156–169, Jan. 2016.

[5] A. Naeem, A. Shabbir, N. U. Hassan, C. Yuen, A. Ahmed, andW. Tushar, “Understanding customer behavior in multi-tierdemandresponse management program,”IEEE Access (Special issue on SmartGrids: A Hub of Interdisciplinary Research), vol. 3, pp. 2613–2625,Nov. 2015.

[6] I. Atzeni, L. G. Ordonez, G. Scutari, D. P. Palomar, andJ. R. Fonollosa,“Demand-side management via distributed energy generation and stor-age optimization,”IEEE Trans. Smart Grid, vol. 4, no. 2, pp. 866–876,June 2013.

[7] Y. Liu, C. Yuen, S. Huang, N. Ul Hassan, X. Wang, and S. Xie,“Peak-to-average ratio constrained demand-side management withconsumer’s

preference in residential smart grid,”IEEE J. Sel. Topics Signal Process.,vol. 8, no. 6, pp. 1084–1097, Dec 2014.

[8] N. U. Hassan, M. A. Pasha, C. Yuen, S. Huang, and X. Wang, “Impact ofscheduling flexibility on demand profile flatness and user inconveniencein residential smart grid system,”Energies, vol. 6, no. 12, pp. 6608–6635, Dec 2012.

[9] T. Jiang, Y. Cao, L. Yu, and Z. Wang, “Load shaping strategy based onenergy storage and dynamic pricing in smart grid,”IEEE Trans. SmartGrid, vol. 5, no. 6, pp. 2868–2876, Nov 2014.

[10] L. Yu, T. Jiang, and Y. Cao, “Energy cost minimization for distributedinternet data centers in smart microgrids considering power outages,”IEEE Trans. Parallel Distrib. Syst., vol. 26, no. 1, pp. 120–130, Jan2015.

[11] W. Tushar, B. Chai, C. Yuen, D. B. Smith, K. L. Wood, Z. Yang, andH. V. Poor, “Three-party energy management with distributed energyresources in smart grid,”IEEE Trans. Ind. Electron., vol. 62, no. 4, pp.2487–2498, Apr. 2015.

[12] Y. Liu, C. Yuen, X. Cao, N. Hassan, and J. Chen, “Design ofa scalablehybrid mac protocol for heterogeneous M2M networks,”IEEE InternetThings J., vol. 1, no. 1, pp. 99–111, Feb 2014.

[13] W. Tushar, J. A. Zhang, D. B. Smith, H. V. Poor, and S. Thi´ebaux,“Prioritizing consumers in smart grid: A game theoretic approach,”IEEETrans. Smart Grid, vol. 5, no. 3, pp. 1429–1438, May 2014.

[14] Y. Cao, T. Jiang, M. He, and J. Zhang, “Device-to-devicecommunica-tions for energy management: A smart grid case,”IEEE J. Sel. AreasCommun., vol. 34, no. 1, pp. 190–201, 2016.

[15] W.-T. Li, C. Yuen, N. U. Hassan, W. Tushar, and C.-K. Wen,“De-mand response management for residential smart grid: From theoryto practice,” IEEE Access (Special issue on Smart Grids: A Hub ofInterdisciplinary Research), vol. 3, pp. 2431–2440, Nov. 2015.